UI-route: An ultra-fast incremental maze routing algorithm
Grid-based maze routing is a fundamental problem in electronic-design-automation (EDA) domain. A core primitive deals with a large query set about route connectivity subject to incremental changes on grid graph. Existing approaches pertain to batch processing, where each route query is independently...
Gespeichert in:
| Veröffentlicht in: | 2014 ACM/IEEE International Workshop on System Level Interconnect Prediction (SLIP) S. 1 - 8 |
|---|---|
| Hauptverfasser: | , , |
| Format: | Tagungsbericht |
| Sprache: | Englisch |
| Veröffentlicht: |
ACM
01.06.2014
|
| Online-Zugang: | Volltext |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| Zusammenfassung: | Grid-based maze routing is a fundamental problem in electronic-design-automation (EDA) domain. A core primitive deals with a large query set about route connectivity subject to incremental changes on grid graph. Existing approaches pertain to batch processing, where each route query is independently and repeatedly solved by a routing procedure. Few researches so far discuss an efficient utilization of search knowledge in incremental fashion, which could dramatically speed up the search. Unfortunately, existing algorithms nearly rely on irregular and highly divergent search space, imposing acceleration challenges on refitting them to incremental version. Consequently, in this paper we present UI-Route, an ultra-fast incremental maze routing algorithm in grid environment. UI-Route is unique in breaking route equivalence, proving that a huge amount of equivalent search efforts can be optimally and incrementally eliminated. Equivalence breaking enables regularized search space, delivering well-tabulated search knowledge through the incremental processing. Moreover UI-Route is largely orthogonal to many applications built upon maze routing and therefore can seamlessly substitute for speedup. Experimental results on a set of modern circuit benchmarks demonstrate that UI-Route achieves prominent speedup over existing algorithms. |
|---|---|
| DOI: | 10.1145/2633948.2633952 |